2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1989 Stephen Deering
5 * Copyright (c) 1992, 1993
6 * The Regents of the University of California. All rights reserved.
8 * This code is derived from software contributed to Berkeley by
9 * Stephen Deering of Stanford University.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
39 * IP multicast forwarding procedures
41 * Written by David Waitzman, BBN Labs, August 1988.
42 * Modified by Steve Deering, Stanford, February 1989.
43 * Modified by Mark J. Steiglitz, Stanford, May, 1991
44 * Modified by Van Jacobson, LBL, January 1993
45 * Modified by Ajit Thyagarajan, PARC, August 1993
46 * Modified by Bill Fenner, PARC, April 1995
47 * Modified by Ahmed Helmy, SGI, June 1996
48 * Modified by George Edmond Eddy (Rusty), ISI, February 1998
49 * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
50 * Modified by Hitoshi Asaeda, WIDE, August 2000
51 * Modified by Pavlin Radoslavov, ICSI, October 2002
53 * MROUTING Revision: 3.5
54 * and PIM-SMv2 and PIM-DM support, advanced API support,
55 * bandwidth metering and signaling
59 * TODO: Prefix functions with ipmf_.
60 * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
61 * domain attachment (if_afdata) so we can track consumers of that service.
62 * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
63 * move it to socket options.
64 * TODO: Cleanup LSRR removal further.
65 * TODO: Push RSVP stubs into raw_ip.c.
66 * TODO: Use bitstring.h for vif set.
67 * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
68 * TODO: Sync ip6_mroute.c with this file.
71 #include <sys/cdefs.h>
72 __FBSDID("$FreeBSD$");
75 #include "opt_mrouting.h"
79 #include <sys/param.h>
80 #include <sys/kernel.h>
81 #include <sys/stddef.h>
82 #include <sys/eventhandler.h>
85 #include <sys/malloc.h>
87 #include <sys/module.h>
89 #include <sys/protosw.h>
90 #include <sys/signalvar.h>
91 #include <sys/socket.h>
92 #include <sys/socketvar.h>
93 #include <sys/sockio.h>
95 #include <sys/sysctl.h>
96 #include <sys/syslog.h>
97 #include <sys/systm.h>
99 #include <sys/counter.h>
102 #include <net/if_var.h>
103 #include <net/netisr.h>
104 #include <net/route.h>
105 #include <net/vnet.h>
107 #include <netinet/in.h>
108 #include <netinet/igmp.h>
109 #include <netinet/in_systm.h>
110 #include <netinet/in_var.h>
111 #include <netinet/ip.h>
112 #include <netinet/ip_encap.h>
113 #include <netinet/ip_mroute.h>
114 #include <netinet/ip_var.h>
115 #include <netinet/ip_options.h>
116 #include <netinet/pim.h>
117 #include <netinet/pim_var.h>
118 #include <netinet/udp.h>
120 #include <machine/in_cksum.h>
123 #define KTR_IPMF KTR_INET
126 #define VIFI_INVALID ((vifi_t) -1)
128 VNET_DEFINE_STATIC(uint32_t, last_tv_sec); /* last time we processed this */
129 #define V_last_tv_sec VNET(last_tv_sec)
131 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
134 * Locking. We use two locks: one for the virtual interface table and
135 * one for the forwarding table. These locks may be nested in which case
136 * the VIF lock must always be taken first. Note that each lock is used
137 * to cover not only the specific data structure but also related data
141 static struct mtx mrouter_mtx;
142 #define MROUTER_LOCK() mtx_lock(&mrouter_mtx)
143 #define MROUTER_UNLOCK() mtx_unlock(&mrouter_mtx)
144 #define MROUTER_LOCK_ASSERT() mtx_assert(&mrouter_mtx, MA_OWNED)
145 #define MROUTER_LOCK_INIT() \
146 mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
147 #define MROUTER_LOCK_DESTROY() mtx_destroy(&mrouter_mtx)
149 static int ip_mrouter_cnt; /* # of vnets with active mrouters */
150 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
152 VNET_PCPUSTAT_DEFINE_STATIC(struct mrtstat, mrtstat);
153 VNET_PCPUSTAT_SYSINIT(mrtstat);
154 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
155 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
156 mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
157 "netinet/ip_mroute.h)");
159 VNET_DEFINE_STATIC(u_long, mfchash);
160 #define V_mfchash VNET(mfchash)
161 #define MFCHASH(a, g) \
162 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
163 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
164 #define MFCHASHSIZE 256
166 static u_long mfchashsize; /* Hash size */
167 VNET_DEFINE_STATIC(u_char *, nexpire); /* 0..mfchashsize-1 */
168 #define V_nexpire VNET(nexpire)
169 VNET_DEFINE_STATIC(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
170 #define V_mfchashtbl VNET(mfchashtbl)
172 static struct mtx mfc_mtx;
173 #define MFC_LOCK() mtx_lock(&mfc_mtx)
174 #define MFC_UNLOCK() mtx_unlock(&mfc_mtx)
175 #define MFC_LOCK_ASSERT() mtx_assert(&mfc_mtx, MA_OWNED)
176 #define MFC_LOCK_INIT() \
177 mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
178 #define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx)
180 VNET_DEFINE_STATIC(vifi_t, numvifs);
181 #define V_numvifs VNET(numvifs)
182 VNET_DEFINE_STATIC(struct vif *, viftable);
183 #define V_viftable VNET(viftable)
185 * No one should be able to "query" this before initialisation happened in
186 * vnet_mroute_init(), so we should still be fine.
188 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_VNET | CTLFLAG_RD,
189 &VNET_NAME(viftable), sizeof(*V_viftable) * MAXVIFS, "S,vif[MAXVIFS]",
190 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
192 static struct mtx vif_mtx;
193 #define VIF_LOCK() mtx_lock(&vif_mtx)
194 #define VIF_UNLOCK() mtx_unlock(&vif_mtx)
195 #define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED)
196 #define VIF_LOCK_INIT() \
197 mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
198 #define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx)
200 static eventhandler_tag if_detach_event_tag = NULL;
202 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
203 #define V_expire_upcalls_ch VNET(expire_upcalls_ch)
205 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
206 #define UPCALL_EXPIRE 6 /* number of timeouts */
209 * Bandwidth meter variables and constants
211 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
213 * Pending timeouts are stored in a hash table, the key being the
214 * expiration time. Periodically, the entries are analysed and processed.
216 #define BW_METER_BUCKETS 1024
217 VNET_DEFINE_STATIC(struct bw_meter **, bw_meter_timers);
218 #define V_bw_meter_timers VNET(bw_meter_timers)
219 VNET_DEFINE_STATIC(struct callout, bw_meter_ch);
220 #define V_bw_meter_ch VNET(bw_meter_ch)
221 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
224 * Pending upcalls are stored in a vector which is flushed when
225 * full, or periodically
227 VNET_DEFINE_STATIC(struct bw_upcall *, bw_upcalls);
228 #define V_bw_upcalls VNET(bw_upcalls)
229 VNET_DEFINE_STATIC(u_int, bw_upcalls_n); /* # of pending upcalls */
230 #define V_bw_upcalls_n VNET(bw_upcalls_n)
231 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
232 #define V_bw_upcalls_ch VNET(bw_upcalls_ch)
234 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
236 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
237 VNET_PCPUSTAT_SYSINIT(pimstat);
238 VNET_PCPUSTAT_SYSUNINIT(pimstat);
240 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
242 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
243 pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
245 static u_long pim_squelch_wholepkt = 0;
246 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
247 &pim_squelch_wholepkt, 0,
248 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
250 static const struct encaptab *pim_encap_cookie;
251 static int pim_encapcheck(const struct mbuf *, int, int, void *);
252 static int pim_input(struct mbuf *, int, int, void *);
254 static const struct encap_config ipv4_encap_cfg = {
255 .proto = IPPROTO_PIM,
256 .min_length = sizeof(struct ip) + PIM_MINLEN,
258 .check = pim_encapcheck,
263 * Note: the PIM Register encapsulation adds the following in front of a
266 * struct pim_encap_hdr {
268 * struct pim_encap_pimhdr pim;
273 struct pim_encap_pimhdr {
277 #define PIM_ENCAP_TTL 64
279 static struct ip pim_encap_iphdr = {
280 #if BYTE_ORDER == LITTLE_ENDIAN
281 sizeof(struct ip) >> 2,
285 sizeof(struct ip) >> 2,
288 sizeof(struct ip), /* total length */
296 static struct pim_encap_pimhdr pim_encap_pimhdr = {
298 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
305 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
306 #define V_reg_vif_num VNET(reg_vif_num)
307 VNET_DEFINE_STATIC(struct ifnet, multicast_register_if);
308 #define V_multicast_register_if VNET(multicast_register_if)
314 static u_long X_ip_mcast_src(int);
315 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
316 struct ip_moptions *);
317 static int X_ip_mrouter_done(void);
318 static int X_ip_mrouter_get(struct socket *, struct sockopt *);
319 static int X_ip_mrouter_set(struct socket *, struct sockopt *);
320 static int X_legal_vif_num(int);
321 static int X_mrt_ioctl(u_long, caddr_t, int);
323 static int add_bw_upcall(struct bw_upcall *);
324 static int add_mfc(struct mfcctl2 *);
325 static int add_vif(struct vifctl *);
326 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
327 static void bw_meter_process(void);
328 static void bw_meter_receive_packet(struct bw_meter *, int,
330 static void bw_upcalls_send(void);
331 static int del_bw_upcall(struct bw_upcall *);
332 static int del_mfc(struct mfcctl2 *);
333 static int del_vif(vifi_t);
334 static int del_vif_locked(vifi_t);
335 static void expire_bw_meter_process(void *);
336 static void expire_bw_upcalls_send(void *);
337 static void expire_mfc(struct mfc *);
338 static void expire_upcalls(void *);
339 static void free_bw_list(struct bw_meter *);
340 static int get_sg_cnt(struct sioc_sg_req *);
341 static int get_vif_cnt(struct sioc_vif_req *);
342 static void if_detached_event(void *, struct ifnet *);
343 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
344 static int ip_mrouter_init(struct socket *, int);
345 static __inline struct mfc *
346 mfc_find(struct in_addr *, struct in_addr *);
347 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
349 pim_register_prepare(struct ip *, struct mbuf *);
350 static int pim_register_send(struct ip *, struct vif *,
351 struct mbuf *, struct mfc *);
352 static int pim_register_send_rp(struct ip *, struct vif *,
353 struct mbuf *, struct mfc *);
354 static int pim_register_send_upcall(struct ip *, struct vif *,
355 struct mbuf *, struct mfc *);
356 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
357 static void send_packet(struct vif *, struct mbuf *);
358 static int set_api_config(uint32_t *);
359 static int set_assert(int);
360 static int socket_send(struct socket *, struct mbuf *,
361 struct sockaddr_in *);
362 static void unschedule_bw_meter(struct bw_meter *);
365 * Kernel multicast forwarding API capabilities and setup.
366 * If more API capabilities are added to the kernel, they should be
367 * recorded in `mrt_api_support'.
369 #define MRT_API_VERSION 0x0305
371 static const int mrt_api_version = MRT_API_VERSION;
372 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
373 MRT_MFC_FLAGS_BORDER_VIF |
376 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
377 #define V_mrt_api_config VNET(mrt_api_config)
378 VNET_DEFINE_STATIC(int, pim_assert_enabled);
379 #define V_pim_assert_enabled VNET(pim_assert_enabled)
380 static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */
383 * Find a route for a given origin IP address and multicast group address.
384 * Statistics must be updated by the caller.
386 static __inline struct mfc *
387 mfc_find(struct in_addr *o, struct in_addr *g)
393 LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
394 if (in_hosteq(rt->mfc_origin, *o) &&
395 in_hosteq(rt->mfc_mcastgrp, *g) &&
396 TAILQ_EMPTY(&rt->mfc_stall))
404 * Handle MRT setsockopt commands to modify the multicast forwarding tables.
407 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
413 struct bw_upcall bw_upcall;
416 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
420 switch (sopt->sopt_name) {
422 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
425 error = ip_mrouter_init(so, optval);
429 error = ip_mrouter_done();
433 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
436 error = add_vif(&vifc);
440 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
443 error = del_vif(vifi);
449 * select data size depending on API version.
451 if (sopt->sopt_name == MRT_ADD_MFC &&
452 V_mrt_api_config & MRT_API_FLAGS_ALL) {
453 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
454 sizeof(struct mfcctl2));
456 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
457 sizeof(struct mfcctl));
458 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
459 sizeof(mfc) - sizeof(struct mfcctl));
463 if (sopt->sopt_name == MRT_ADD_MFC)
464 error = add_mfc(&mfc);
466 error = del_mfc(&mfc);
470 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
477 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
479 error = set_api_config(&i);
481 error = sooptcopyout(sopt, &i, sizeof i);
484 case MRT_ADD_BW_UPCALL:
485 case MRT_DEL_BW_UPCALL:
486 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
490 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
491 error = add_bw_upcall(&bw_upcall);
493 error = del_bw_upcall(&bw_upcall);
504 * Handle MRT getsockopt commands
507 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
511 switch (sopt->sopt_name) {
513 error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
517 error = sooptcopyout(sopt, &V_pim_assert_enabled,
518 sizeof V_pim_assert_enabled);
521 case MRT_API_SUPPORT:
522 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
526 error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
537 * Handle ioctl commands to obtain information from the cache
540 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
545 * Currently the only function calling this ioctl routine is rtioctl_fib().
546 * Typically, only root can create the raw socket in order to execute
547 * this ioctl method, however the request might be coming from a prison
549 error = priv_check(curthread, PRIV_NETINET_MROUTE);
553 case (SIOCGETVIFCNT):
554 error = get_vif_cnt((struct sioc_vif_req *)data);
558 error = get_sg_cnt((struct sioc_sg_req *)data);
569 * returns the packet, byte, rpf-failure count for the source group provided
572 get_sg_cnt(struct sioc_sg_req *req)
577 rt = mfc_find(&req->src, &req->grp);
580 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
581 return EADDRNOTAVAIL;
583 req->pktcnt = rt->mfc_pkt_cnt;
584 req->bytecnt = rt->mfc_byte_cnt;
585 req->wrong_if = rt->mfc_wrong_if;
591 * returns the input and output packet and byte counts on the vif provided
594 get_vif_cnt(struct sioc_vif_req *req)
596 vifi_t vifi = req->vifi;
599 if (vifi >= V_numvifs) {
604 req->icount = V_viftable[vifi].v_pkt_in;
605 req->ocount = V_viftable[vifi].v_pkt_out;
606 req->ibytes = V_viftable[vifi].v_bytes_in;
607 req->obytes = V_viftable[vifi].v_bytes_out;
614 if_detached_event(void *arg __unused, struct ifnet *ifp)
621 if (V_ip_mrouter == NULL) {
630 * Tear down multicast forwarder state associated with this ifnet.
631 * 1. Walk the vif list, matching vifs against this ifnet.
632 * 2. Walk the multicast forwarding cache (mfc) looking for
633 * inner matches with this vif's index.
634 * 3. Expire any matching multicast forwarding cache entries.
635 * 4. Free vif state. This should disable ALLMULTI on the interface.
637 for (vifi = 0; vifi < V_numvifs; vifi++) {
638 if (V_viftable[vifi].v_ifp != ifp)
640 for (i = 0; i < mfchashsize; i++) {
641 struct mfc *rt, *nrt;
643 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
644 if (rt->mfc_parent == vifi) {
649 del_vif_locked(vifi);
659 * Enable multicast forwarding.
662 ip_mrouter_init(struct socket *so, int version)
665 CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
666 so->so_type, so->so_proto->pr_protocol);
668 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
676 if (ip_mrouter_unloading) {
681 if (V_ip_mrouter != NULL) {
686 V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
689 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
691 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
693 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
701 CTR1(KTR_IPMF, "%s: done", __func__);
707 * Disable multicast forwarding.
710 X_ip_mrouter_done(void)
718 if (V_ip_mrouter == NULL) {
724 * Detach/disable hooks to the reset of the system.
728 V_mrt_api_config = 0;
733 * For each phyint in use, disable promiscuous reception of all IP
736 for (vifi = 0; vifi < V_numvifs; vifi++) {
737 if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
738 !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
739 ifp = V_viftable[vifi].v_ifp;
743 bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
745 V_pim_assert_enabled = 0;
749 callout_stop(&V_expire_upcalls_ch);
750 callout_stop(&V_bw_upcalls_ch);
751 callout_stop(&V_bw_meter_ch);
756 * Free all multicast forwarding cache entries.
757 * Do not use hashdestroy(), as we must perform other cleanup.
759 for (i = 0; i < mfchashsize; i++) {
760 struct mfc *rt, *nrt;
762 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
766 free(V_mfchashtbl, M_MRTABLE);
769 bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
772 bzero(V_bw_meter_timers, BW_METER_BUCKETS * sizeof(*V_bw_meter_timers));
776 V_reg_vif_num = VIFI_INVALID;
780 CTR1(KTR_IPMF, "%s: done", __func__);
786 * Set PIM assert processing global
791 if ((i != 1) && (i != 0))
794 V_pim_assert_enabled = i;
800 * Configure API capabilities
803 set_api_config(uint32_t *apival)
808 * We can set the API capabilities only if it is the first operation
809 * after MRT_INIT. I.e.:
810 * - there are no vifs installed
811 * - pim_assert is not enabled
812 * - the MFC table is empty
818 if (V_pim_assert_enabled) {
825 for (i = 0; i < mfchashsize; i++) {
826 if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
835 V_mrt_api_config = *apival & mrt_api_support;
836 *apival = V_mrt_api_config;
842 * Add a vif to the vif table
845 add_vif(struct vifctl *vifcp)
847 struct vif *vifp = V_viftable + vifcp->vifc_vifi;
848 struct sockaddr_in sin = {sizeof sin, AF_INET};
854 if (vifcp->vifc_vifi >= MAXVIFS) {
858 /* rate limiting is no longer supported by this code */
859 if (vifcp->vifc_rate_limit != 0) {
860 log(LOG_ERR, "rate limiting is no longer supported\n");
864 if (!in_nullhost(vifp->v_lcl_addr)) {
868 if (in_nullhost(vifcp->vifc_lcl_addr)) {
870 return EADDRNOTAVAIL;
873 /* Find the interface with an address in AF_INET family */
874 if (vifcp->vifc_flags & VIFF_REGISTER) {
876 * XXX: Because VIFF_REGISTER does not really need a valid
877 * local interface (e.g. it could be 127.0.0.2), we don't
882 struct epoch_tracker et;
884 sin.sin_addr = vifcp->vifc_lcl_addr;
886 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
890 return EADDRNOTAVAIL;
893 /* XXX FIXME we need to take a ref on ifp and cleanup properly! */
897 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
898 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
901 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
902 ifp = &V_multicast_register_if;
903 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
904 if (V_reg_vif_num == VIFI_INVALID) {
905 if_initname(&V_multicast_register_if, "register_vif", 0);
906 V_multicast_register_if.if_flags = IFF_LOOPBACK;
907 V_reg_vif_num = vifcp->vifc_vifi;
909 } else { /* Make sure the interface supports multicast */
910 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
915 /* Enable promiscuous reception of all IP multicasts from the if */
916 error = if_allmulti(ifp, 1);
923 vifp->v_flags = vifcp->vifc_flags;
924 vifp->v_threshold = vifcp->vifc_threshold;
925 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
926 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
928 /* initialize per vif pkt counters */
931 vifp->v_bytes_in = 0;
932 vifp->v_bytes_out = 0;
934 /* Adjust numvifs up if the vifi is higher than numvifs */
935 if (V_numvifs <= vifcp->vifc_vifi)
936 V_numvifs = vifcp->vifc_vifi + 1;
940 CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
941 (int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
942 (int)vifcp->vifc_threshold);
948 * Delete a vif from the vif table
951 del_vif_locked(vifi_t vifi)
957 if (vifi >= V_numvifs) {
960 vifp = &V_viftable[vifi];
961 if (in_nullhost(vifp->v_lcl_addr)) {
962 return EADDRNOTAVAIL;
965 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
966 if_allmulti(vifp->v_ifp, 0);
968 if (vifp->v_flags & VIFF_REGISTER)
969 V_reg_vif_num = VIFI_INVALID;
971 bzero((caddr_t)vifp, sizeof (*vifp));
973 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
975 /* Adjust numvifs down */
976 for (vifi = V_numvifs; vifi > 0; vifi--)
977 if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
990 cc = del_vif_locked(vifi);
997 * update an mfc entry without resetting counters and S,G addresses.
1000 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1004 rt->mfc_parent = mfccp->mfcc_parent;
1005 for (i = 0; i < V_numvifs; i++) {
1006 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1007 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1010 /* set the RP address */
1011 if (V_mrt_api_config & MRT_MFC_RP)
1012 rt->mfc_rp = mfccp->mfcc_rp;
1014 rt->mfc_rp.s_addr = INADDR_ANY;
1018 * fully initialize an mfc entry from the parameter.
1021 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1023 rt->mfc_origin = mfccp->mfcc_origin;
1024 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1026 update_mfc_params(rt, mfccp);
1028 /* initialize pkt counters per src-grp */
1029 rt->mfc_pkt_cnt = 0;
1030 rt->mfc_byte_cnt = 0;
1031 rt->mfc_wrong_if = 0;
1032 timevalclear(&rt->mfc_last_assert);
1036 expire_mfc(struct mfc *rt)
1038 struct rtdetq *rte, *nrte;
1042 free_bw_list(rt->mfc_bw_meter);
1044 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1046 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1047 free(rte, M_MRTABLE);
1050 LIST_REMOVE(rt, mfc_hash);
1051 free(rt, M_MRTABLE);
1058 add_mfc(struct mfcctl2 *mfccp)
1061 struct rtdetq *rte, *nrte;
1068 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1070 /* If an entry already exists, just update the fields */
1072 CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1073 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1074 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1075 mfccp->mfcc_parent);
1076 update_mfc_params(rt, mfccp);
1083 * Find the entry for which the upcall was made and update
1086 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1087 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1088 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1089 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1090 !TAILQ_EMPTY(&rt->mfc_stall)) {
1092 "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1093 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1094 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1096 TAILQ_FIRST(&rt->mfc_stall));
1098 CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1100 init_mfc_params(rt, mfccp);
1101 rt->mfc_expire = 0; /* Don't clean this guy up */
1104 /* Free queued packets, but attempt to forward them first. */
1105 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1106 if (rte->ifp != NULL)
1107 ip_mdq(rte->m, rte->ifp, rt, -1);
1109 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1111 free(rte, M_MRTABLE);
1117 * It is possible that an entry is being inserted without an upcall
1120 CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1121 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1122 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1123 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1124 init_mfc_params(rt, mfccp);
1132 if (rt == NULL) { /* no upcall, so make a new entry */
1133 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1140 init_mfc_params(rt, mfccp);
1141 TAILQ_INIT(&rt->mfc_stall);
1145 rt->mfc_bw_meter = NULL;
1147 /* insert new entry at head of hash chain */
1148 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1159 * Delete an mfc entry
1162 del_mfc(struct mfcctl2 *mfccp)
1164 struct in_addr origin;
1165 struct in_addr mcastgrp;
1168 origin = mfccp->mfcc_origin;
1169 mcastgrp = mfccp->mfcc_mcastgrp;
1171 CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1172 ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1176 rt = mfc_find(&origin, &mcastgrp);
1179 return EADDRNOTAVAIL;
1183 * free the bw_meter entries
1185 free_bw_list(rt->mfc_bw_meter);
1186 rt->mfc_bw_meter = NULL;
1188 LIST_REMOVE(rt, mfc_hash);
1189 free(rt, M_MRTABLE);
1197 * Send a message to the routing daemon on the multicast routing socket.
1200 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1203 SOCKBUF_LOCK(&s->so_rcv);
1204 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1206 sorwakeup_locked(s);
1209 SOCKBUF_UNLOCK(&s->so_rcv);
1216 * IP multicast forwarding function. This function assumes that the packet
1217 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1218 * pointed to by "ifp", and the packet is to be relayed to other networks
1219 * that have members of the packet's destination IP multicast group.
1221 * The packet is returned unscathed to the caller, unless it is
1222 * erroneous, in which case a non-zero return value tells the caller to
1226 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1229 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1230 struct ip_moptions *imo)
1236 CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1237 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1239 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1240 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1242 * Packet arrived via a physical interface or
1243 * an encapsulated tunnel or a register_vif.
1247 * Packet arrived through a source-route tunnel.
1248 * Source-route tunnels are no longer supported.
1255 if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1256 if (ip->ip_ttl < MAXTTL)
1257 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1258 error = ip_mdq(m, ifp, NULL, vifi);
1265 * Don't forward a packet with time-to-live of zero or one,
1266 * or a packet destined to a local-only group.
1268 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1275 * Determine forwarding vifs from the forwarding cache table
1277 MRTSTAT_INC(mrts_mfc_lookups);
1278 rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1280 /* Entry exists, so forward if necessary */
1282 error = ip_mdq(m, ifp, rt, -1);
1288 * If we don't have a route for packet's origin,
1289 * Make a copy of the packet & send message to routing daemon
1295 int hlen = ip->ip_hl << 2;
1297 MRTSTAT_INC(mrts_mfc_misses);
1298 MRTSTAT_INC(mrts_no_route);
1299 CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1300 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1303 * Allocate mbufs early so that we don't do extra work if we are
1304 * just going to fail anyway. Make sure to pullup the header so
1305 * that other people can't step on it.
1307 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1315 mb0 = m_copypacket(m, M_NOWAIT);
1316 if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1317 mb0 = m_pullup(mb0, hlen);
1319 free(rte, M_MRTABLE);
1325 /* is there an upcall waiting for this flow ? */
1326 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1327 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1328 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1329 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1330 !TAILQ_EMPTY(&rt->mfc_stall))
1337 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1341 * Locate the vifi for the incoming interface for this packet.
1342 * If none found, drop packet.
1344 for (vifi = 0; vifi < V_numvifs &&
1345 V_viftable[vifi].v_ifp != ifp; vifi++)
1347 if (vifi >= V_numvifs) /* vif not found, drop packet */
1350 /* no upcall, so make a new entry */
1351 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1355 /* Make a copy of the header to send to the user level process */
1356 mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1361 * Send message to routing daemon to install
1362 * a route into the kernel table
1365 im = mtod(mm, struct igmpmsg *);
1366 im->im_msgtype = IGMPMSG_NOCACHE;
1370 MRTSTAT_INC(mrts_upcalls);
1372 k_igmpsrc.sin_addr = ip->ip_src;
1373 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1374 CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1375 MRTSTAT_INC(mrts_upq_sockfull);
1377 free(rt, M_MRTABLE);
1379 free(rte, M_MRTABLE);
1386 /* insert new entry at head of hash chain */
1387 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1388 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1389 rt->mfc_expire = UPCALL_EXPIRE;
1391 for (i = 0; i < V_numvifs; i++) {
1392 rt->mfc_ttls[i] = 0;
1393 rt->mfc_flags[i] = 0;
1395 rt->mfc_parent = -1;
1397 /* clear the RP address */
1398 rt->mfc_rp.s_addr = INADDR_ANY;
1399 rt->mfc_bw_meter = NULL;
1401 /* initialize pkt counters per src-grp */
1402 rt->mfc_pkt_cnt = 0;
1403 rt->mfc_byte_cnt = 0;
1404 rt->mfc_wrong_if = 0;
1405 timevalclear(&rt->mfc_last_assert);
1407 TAILQ_INIT(&rt->mfc_stall);
1410 /* link into table */
1411 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1412 TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1416 /* determine if queue has overflowed */
1417 if (rt->mfc_nstall > MAX_UPQ) {
1418 MRTSTAT_INC(mrts_upq_ovflw);
1420 free(rte, M_MRTABLE);
1426 TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1441 * Clean up the cache entry if upcall is not serviced
1444 expire_upcalls(void *arg)
1448 CURVNET_SET((struct vnet *) arg);
1452 for (i = 0; i < mfchashsize; i++) {
1453 struct mfc *rt, *nrt;
1455 if (V_nexpire[i] == 0)
1458 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1459 if (TAILQ_EMPTY(&rt->mfc_stall))
1462 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1466 * free the bw_meter entries
1468 while (rt->mfc_bw_meter != NULL) {
1469 struct bw_meter *x = rt->mfc_bw_meter;
1471 rt->mfc_bw_meter = x->bm_mfc_next;
1475 MRTSTAT_INC(mrts_cache_cleanups);
1476 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1477 (u_long)ntohl(rt->mfc_origin.s_addr),
1478 (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1486 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1493 * Packet forwarding routine once entry in the cache is made
1496 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1498 struct ip *ip = mtod(m, struct ip *);
1500 int plen = ntohs(ip->ip_len);
1505 * If xmt_vif is not -1, send on only the requested vif.
1507 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1509 if (xmt_vif < V_numvifs) {
1510 if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1511 pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1513 phyint_send(ip, V_viftable + xmt_vif, m);
1518 * Don't forward if it didn't arrive from the parent vif for its origin.
1520 vifi = rt->mfc_parent;
1521 if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1522 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1523 __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1524 MRTSTAT_INC(mrts_wrong_if);
1527 * If we are doing PIM assert processing, send a message
1528 * to the routing daemon.
1530 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1531 * can complete the SPT switch, regardless of the type
1532 * of the iif (broadcast media, GRE tunnel, etc).
1534 if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1535 V_viftable[vifi].v_ifp) {
1537 if (ifp == &V_multicast_register_if)
1538 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1540 /* Get vifi for the incoming packet */
1541 for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1544 if (vifi >= V_numvifs)
1545 return 0; /* The iif is not found: ignore the packet. */
1547 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1548 return 0; /* WRONGVIF disabled: ignore the packet */
1550 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1551 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1553 int hlen = ip->ip_hl << 2;
1554 struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1556 if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1557 mm = m_pullup(mm, hlen);
1561 im = mtod(mm, struct igmpmsg *);
1562 im->im_msgtype = IGMPMSG_WRONGVIF;
1566 MRTSTAT_INC(mrts_upcalls);
1568 k_igmpsrc.sin_addr = im->im_src;
1569 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1570 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1571 MRTSTAT_INC(mrts_upq_sockfull);
1580 /* If I sourced this packet, it counts as output, else it was input. */
1581 if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1582 V_viftable[vifi].v_pkt_out++;
1583 V_viftable[vifi].v_bytes_out += plen;
1585 V_viftable[vifi].v_pkt_in++;
1586 V_viftable[vifi].v_bytes_in += plen;
1589 rt->mfc_byte_cnt += plen;
1592 * For each vif, decide if a copy of the packet should be forwarded.
1594 * - the ttl exceeds the vif's threshold
1595 * - there are group members downstream on interface
1597 for (vifi = 0; vifi < V_numvifs; vifi++)
1598 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1599 V_viftable[vifi].v_pkt_out++;
1600 V_viftable[vifi].v_bytes_out += plen;
1601 if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1602 pim_register_send(ip, V_viftable + vifi, m, rt);
1604 phyint_send(ip, V_viftable + vifi, m);
1608 * Perform upcall-related bw measuring.
1610 if (rt->mfc_bw_meter != NULL) {
1616 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1617 bw_meter_receive_packet(x, plen, &now);
1624 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1627 X_legal_vif_num(int vif)
1636 if (vif < V_numvifs)
1644 * Return the local address used by this vif
1647 X_ip_mcast_src(int vifi)
1656 if (vifi < V_numvifs)
1657 addr = V_viftable[vifi].v_lcl_addr.s_addr;
1664 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1666 struct mbuf *mb_copy;
1667 int hlen = ip->ip_hl << 2;
1672 * Make a new reference to the packet; make sure that
1673 * the IP header is actually copied, not just referenced,
1674 * so that ip_output() only scribbles on the copy.
1676 mb_copy = m_copypacket(m, M_NOWAIT);
1677 if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1678 mb_copy = m_pullup(mb_copy, hlen);
1679 if (mb_copy == NULL)
1682 send_packet(vifp, mb_copy);
1686 send_packet(struct vif *vifp, struct mbuf *m)
1688 struct ip_moptions imo;
1693 imo.imo_multicast_ifp = vifp->v_ifp;
1694 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1695 imo.imo_multicast_loop = 1;
1696 imo.imo_multicast_vif = -1;
1697 STAILQ_INIT(&imo.imo_head);
1700 * Re-entrancy should not be a problem here, because
1701 * the packets that we send out and are looped back at us
1702 * should get rejected because they appear to come from
1703 * the loopback interface, thus preventing looping.
1705 error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1706 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1707 (ptrdiff_t)(vifp - V_viftable), error);
1711 * Stubs for old RSVP socket shim implementation.
1715 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1718 return (EOPNOTSUPP);
1722 X_ip_rsvp_force_done(struct socket *so __unused)
1728 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1736 return (IPPROTO_DONE);
1740 * Code for bandwidth monitors
1744 * Define common interface for timeval-related methods
1746 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1747 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1748 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1751 compute_bw_meter_flags(struct bw_upcall *req)
1755 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1756 flags |= BW_METER_UNIT_PACKETS;
1757 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1758 flags |= BW_METER_UNIT_BYTES;
1759 if (req->bu_flags & BW_UPCALL_GEQ)
1760 flags |= BW_METER_GEQ;
1761 if (req->bu_flags & BW_UPCALL_LEQ)
1762 flags |= BW_METER_LEQ;
1768 * Add a bw_meter entry
1771 add_bw_upcall(struct bw_upcall *req)
1774 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1775 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1780 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1783 /* Test if the flags are valid */
1784 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1786 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1788 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1789 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1792 /* Test if the threshold time interval is valid */
1793 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1796 flags = compute_bw_meter_flags(req);
1799 * Find if we have already same bw_meter entry
1802 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1805 return EADDRNOTAVAIL;
1807 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1808 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1809 &req->bu_threshold.b_time, ==)) &&
1810 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1811 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1812 (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
1814 return 0; /* XXX Already installed */
1818 /* Allocate the new bw_meter entry */
1819 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1825 /* Set the new bw_meter entry */
1826 x->bm_threshold.b_time = req->bu_threshold.b_time;
1828 x->bm_start_time = now;
1829 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1830 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1831 x->bm_measured.b_packets = 0;
1832 x->bm_measured.b_bytes = 0;
1833 x->bm_flags = flags;
1834 x->bm_time_next = NULL;
1835 x->bm_time_hash = BW_METER_BUCKETS;
1837 /* Add the new bw_meter entry to the front of entries for this MFC */
1839 x->bm_mfc_next = mfc->mfc_bw_meter;
1840 mfc->mfc_bw_meter = x;
1841 schedule_bw_meter(x, &now);
1848 free_bw_list(struct bw_meter *list)
1850 while (list != NULL) {
1851 struct bw_meter *x = list;
1853 list = list->bm_mfc_next;
1854 unschedule_bw_meter(x);
1860 * Delete one or multiple bw_meter entries
1863 del_bw_upcall(struct bw_upcall *req)
1868 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1873 /* Find the corresponding MFC entry */
1874 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1877 return EADDRNOTAVAIL;
1878 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1880 * Delete all bw_meter entries for this mfc
1882 struct bw_meter *list;
1884 list = mfc->mfc_bw_meter;
1885 mfc->mfc_bw_meter = NULL;
1889 } else { /* Delete a single bw_meter entry */
1890 struct bw_meter *prev;
1893 flags = compute_bw_meter_flags(req);
1895 /* Find the bw_meter entry to delete */
1896 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1897 prev = x, x = x->bm_mfc_next) {
1898 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1899 &req->bu_threshold.b_time, ==)) &&
1900 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1901 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1902 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
1905 if (x != NULL) { /* Delete entry from the list for this MFC */
1907 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
1909 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1911 unschedule_bw_meter(x);
1913 /* Free the bw_meter entry */
1925 * Perform bandwidth measurement processing that may result in an upcall
1928 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1930 struct timeval delta;
1935 BW_TIMEVALDECR(&delta, &x->bm_start_time);
1937 if (x->bm_flags & BW_METER_GEQ) {
1939 * Processing for ">=" type of bw_meter entry
1941 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1942 /* Reset the bw_meter entry */
1943 x->bm_start_time = *nowp;
1944 x->bm_measured.b_packets = 0;
1945 x->bm_measured.b_bytes = 0;
1946 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1949 /* Record that a packet is received */
1950 x->bm_measured.b_packets++;
1951 x->bm_measured.b_bytes += plen;
1954 * Test if we should deliver an upcall
1956 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1957 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1958 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1959 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1960 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1961 /* Prepare an upcall for delivery */
1962 bw_meter_prepare_upcall(x, nowp);
1963 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1966 } else if (x->bm_flags & BW_METER_LEQ) {
1968 * Processing for "<=" type of bw_meter entry
1970 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1972 * We are behind time with the multicast forwarding table
1973 * scanning for "<=" type of bw_meter entries, so test now
1974 * if we should deliver an upcall.
1976 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1977 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1978 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1979 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1980 /* Prepare an upcall for delivery */
1981 bw_meter_prepare_upcall(x, nowp);
1983 /* Reschedule the bw_meter entry */
1984 unschedule_bw_meter(x);
1985 schedule_bw_meter(x, nowp);
1988 /* Record that a packet is received */
1989 x->bm_measured.b_packets++;
1990 x->bm_measured.b_bytes += plen;
1993 * Test if we should restart the measuring interval
1995 if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1996 x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1997 (x->bm_flags & BW_METER_UNIT_BYTES &&
1998 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1999 /* Don't restart the measuring interval */
2001 /* Do restart the measuring interval */
2003 * XXX: note that we don't unschedule and schedule, because this
2004 * might be too much overhead per packet. Instead, when we process
2005 * all entries for a given timer hash bin, we check whether it is
2006 * really a timeout. If not, we reschedule at that time.
2008 x->bm_start_time = *nowp;
2009 x->bm_measured.b_packets = 0;
2010 x->bm_measured.b_bytes = 0;
2011 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2017 * Prepare a bandwidth-related upcall
2020 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2022 struct timeval delta;
2023 struct bw_upcall *u;
2028 * Compute the measured time interval
2031 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2034 * If there are too many pending upcalls, deliver them now
2036 if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2040 * Set the bw_upcall entry
2042 u = &V_bw_upcalls[V_bw_upcalls_n++];
2043 u->bu_src = x->bm_mfc->mfc_origin;
2044 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2045 u->bu_threshold.b_time = x->bm_threshold.b_time;
2046 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2047 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2048 u->bu_measured.b_time = delta;
2049 u->bu_measured.b_packets = x->bm_measured.b_packets;
2050 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2052 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2053 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2054 if (x->bm_flags & BW_METER_UNIT_BYTES)
2055 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2056 if (x->bm_flags & BW_METER_GEQ)
2057 u->bu_flags |= BW_UPCALL_GEQ;
2058 if (x->bm_flags & BW_METER_LEQ)
2059 u->bu_flags |= BW_UPCALL_LEQ;
2063 * Send the pending bandwidth-related upcalls
2066 bw_upcalls_send(void)
2069 int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2070 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2071 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2073 IGMPMSG_BW_UPCALL,/* im_msgtype */
2078 { 0 } }; /* im_dst */
2082 if (V_bw_upcalls_n == 0)
2083 return; /* No pending upcalls */
2088 * Allocate a new mbuf, initialize it with the header and
2089 * the payload for the pending calls.
2091 m = m_gethdr(M_NOWAIT, MT_DATA);
2093 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2097 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2098 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2102 * XXX do we need to set the address in k_igmpsrc ?
2104 MRTSTAT_INC(mrts_upcalls);
2105 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2106 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2107 MRTSTAT_INC(mrts_upq_sockfull);
2112 * Compute the timeout hash value for the bw_meter entries
2114 #define BW_METER_TIMEHASH(bw_meter, hash) \
2116 struct timeval next_timeval = (bw_meter)->bm_start_time; \
2118 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2119 (hash) = next_timeval.tv_sec; \
2120 if (next_timeval.tv_usec) \
2121 (hash)++; /* XXX: make sure we don't timeout early */ \
2122 (hash) %= BW_METER_BUCKETS; \
2126 * Schedule a timer to process periodically bw_meter entry of type "<="
2127 * by linking the entry in the proper hash bucket.
2130 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2136 if (!(x->bm_flags & BW_METER_LEQ))
2137 return; /* XXX: we schedule timers only for "<=" entries */
2140 * Reset the bw_meter entry
2142 x->bm_start_time = *nowp;
2143 x->bm_measured.b_packets = 0;
2144 x->bm_measured.b_bytes = 0;
2145 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2148 * Compute the timeout hash value and insert the entry
2150 BW_METER_TIMEHASH(x, time_hash);
2151 x->bm_time_next = V_bw_meter_timers[time_hash];
2152 V_bw_meter_timers[time_hash] = x;
2153 x->bm_time_hash = time_hash;
2157 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2158 * by removing the entry from the proper hash bucket.
2161 unschedule_bw_meter(struct bw_meter *x)
2164 struct bw_meter *prev, *tmp;
2168 if (!(x->bm_flags & BW_METER_LEQ))
2169 return; /* XXX: we schedule timers only for "<=" entries */
2172 * Compute the timeout hash value and delete the entry
2174 time_hash = x->bm_time_hash;
2175 if (time_hash >= BW_METER_BUCKETS)
2176 return; /* Entry was not scheduled */
2178 for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2179 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2184 panic("unschedule_bw_meter: bw_meter entry not found");
2187 prev->bm_time_next = x->bm_time_next;
2189 V_bw_meter_timers[time_hash] = x->bm_time_next;
2191 x->bm_time_next = NULL;
2192 x->bm_time_hash = BW_METER_BUCKETS;
2197 * Process all "<=" type of bw_meter that should be processed now,
2198 * and for each entry prepare an upcall if necessary. Each processed
2199 * entry is rescheduled again for the (periodic) processing.
2201 * This is run periodically (once per second normally). On each round,
2202 * all the potentially matching entries are in the hash slot that we are
2210 struct timeval now, process_endtime;
2213 if (V_last_tv_sec == now.tv_sec)
2214 return; /* nothing to do */
2216 loops = now.tv_sec - V_last_tv_sec;
2217 V_last_tv_sec = now.tv_sec;
2218 if (loops > BW_METER_BUCKETS)
2219 loops = BW_METER_BUCKETS;
2223 * Process all bins of bw_meter entries from the one after the last
2224 * processed to the current one. On entry, i points to the last bucket
2225 * visited, so we need to increment i at the beginning of the loop.
2227 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2228 struct bw_meter *x, *tmp_list;
2230 if (++i >= BW_METER_BUCKETS)
2233 /* Disconnect the list of bw_meter entries from the bin */
2234 tmp_list = V_bw_meter_timers[i];
2235 V_bw_meter_timers[i] = NULL;
2237 /* Process the list of bw_meter entries */
2238 while (tmp_list != NULL) {
2240 tmp_list = tmp_list->bm_time_next;
2242 /* Test if the time interval is over */
2243 process_endtime = x->bm_start_time;
2244 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2245 if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2246 /* Not yet: reschedule, but don't reset */
2249 BW_METER_TIMEHASH(x, time_hash);
2250 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2252 * XXX: somehow the bin processing is a bit ahead of time.
2253 * Put the entry in the next bin.
2255 if (++time_hash >= BW_METER_BUCKETS)
2258 x->bm_time_next = V_bw_meter_timers[time_hash];
2259 V_bw_meter_timers[time_hash] = x;
2260 x->bm_time_hash = time_hash;
2266 * Test if we should deliver an upcall
2268 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2269 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2270 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2271 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2272 /* Prepare an upcall for delivery */
2273 bw_meter_prepare_upcall(x, &now);
2277 * Reschedule for next processing
2279 schedule_bw_meter(x, &now);
2283 /* Send all upcalls that are pending delivery */
2290 * A periodic function for sending all upcalls that are pending delivery
2293 expire_bw_upcalls_send(void *arg)
2295 CURVNET_SET((struct vnet *) arg);
2301 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2307 * A periodic function for periodic scanning of the multicast forwarding
2308 * table for processing all "<=" bw_meter entries.
2311 expire_bw_meter_process(void *arg)
2313 CURVNET_SET((struct vnet *) arg);
2315 if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2318 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2324 * End of bandwidth monitoring code
2328 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2332 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2335 struct mbuf *mb_copy, *mm;
2338 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2339 * rendezvous point was unspecified, and we were told not to.
2341 if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2342 in_nullhost(rt->mfc_rp))
2345 mb_copy = pim_register_prepare(ip, m);
2346 if (mb_copy == NULL)
2350 * Send all the fragments. Note that the mbuf for each fragment
2351 * is freed by the sending machinery.
2353 for (mm = mb_copy; mm; mm = mb_copy) {
2354 mb_copy = mm->m_nextpkt;
2356 mm = m_pullup(mm, sizeof(struct ip));
2358 ip = mtod(mm, struct ip *);
2359 if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2360 pim_register_send_rp(ip, vifp, mm, rt);
2362 pim_register_send_upcall(ip, vifp, mm, rt);
2371 * Return a copy of the data packet that is ready for PIM Register
2373 * XXX: Note that in the returned copy the IP header is a valid one.
2375 static struct mbuf *
2376 pim_register_prepare(struct ip *ip, struct mbuf *m)
2378 struct mbuf *mb_copy = NULL;
2381 /* Take care of delayed checksums */
2382 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2383 in_delayed_cksum(m);
2384 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2388 * Copy the old packet & pullup its IP header into the
2389 * new mbuf so we can modify it.
2391 mb_copy = m_copypacket(m, M_NOWAIT);
2392 if (mb_copy == NULL)
2394 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2395 if (mb_copy == NULL)
2398 /* take care of the TTL */
2399 ip = mtod(mb_copy, struct ip *);
2402 /* Compute the MTU after the PIM Register encapsulation */
2403 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2405 if (ntohs(ip->ip_len) <= mtu) {
2406 /* Turn the IP header into a valid one */
2408 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2410 /* Fragment the packet */
2411 mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2412 if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2421 * Send an upcall with the data packet to the user-level process.
2424 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2425 struct mbuf *mb_copy, struct mfc *rt)
2427 struct mbuf *mb_first;
2428 int len = ntohs(ip->ip_len);
2430 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2435 * Add a new mbuf with an upcall header
2437 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2438 if (mb_first == NULL) {
2442 mb_first->m_data += max_linkhdr;
2443 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2444 mb_first->m_len = sizeof(struct igmpmsg);
2445 mb_first->m_next = mb_copy;
2447 /* Send message to routing daemon */
2448 im = mtod(mb_first, struct igmpmsg *);
2449 im->im_msgtype = IGMPMSG_WHOLEPKT;
2451 im->im_vif = vifp - V_viftable;
2452 im->im_src = ip->ip_src;
2453 im->im_dst = ip->ip_dst;
2455 k_igmpsrc.sin_addr = ip->ip_src;
2457 MRTSTAT_INC(mrts_upcalls);
2459 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2460 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2461 MRTSTAT_INC(mrts_upq_sockfull);
2465 /* Keep statistics */
2466 PIMSTAT_INC(pims_snd_registers_msgs);
2467 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2473 * Encapsulate the data packet in PIM Register message and send it to the RP.
2476 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2479 struct mbuf *mb_first;
2480 struct ip *ip_outer;
2481 struct pim_encap_pimhdr *pimhdr;
2482 int len = ntohs(ip->ip_len);
2483 vifi_t vifi = rt->mfc_parent;
2487 if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2489 return EADDRNOTAVAIL; /* The iif vif is invalid */
2493 * Add a new mbuf with the encapsulating header
2495 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2496 if (mb_first == NULL) {
2500 mb_first->m_data += max_linkhdr;
2501 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2502 mb_first->m_next = mb_copy;
2504 mb_first->m_pkthdr.len = len + mb_first->m_len;
2507 * Fill in the encapsulating IP and PIM header
2509 ip_outer = mtod(mb_first, struct ip *);
2510 *ip_outer = pim_encap_iphdr;
2511 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2512 sizeof(pim_encap_pimhdr));
2513 ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2514 ip_outer->ip_dst = rt->mfc_rp;
2516 * Copy the inner header TOS to the outer header, and take care of the
2519 ip_outer->ip_tos = ip->ip_tos;
2520 if (ip->ip_off & htons(IP_DF))
2521 ip_outer->ip_off |= htons(IP_DF);
2522 ip_fillid(ip_outer);
2523 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2524 + sizeof(pim_encap_iphdr));
2525 *pimhdr = pim_encap_pimhdr;
2526 /* If the iif crosses a border, set the Border-bit */
2527 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2528 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2530 mb_first->m_data += sizeof(pim_encap_iphdr);
2531 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2532 mb_first->m_data -= sizeof(pim_encap_iphdr);
2534 send_packet(vifp, mb_first);
2536 /* Keep statistics */
2537 PIMSTAT_INC(pims_snd_registers_msgs);
2538 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2544 * pim_encapcheck() is called by the encap4_input() path at runtime to
2545 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2549 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2550 int proto __unused, void *arg __unused)
2553 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2554 return (8); /* claim the datagram. */
2558 * PIM-SMv2 and PIM-DM messages processing.
2559 * Receives and verifies the PIM control messages, and passes them
2560 * up to the listening socket, using rip_input().
2561 * The only message with special processing is the PIM_REGISTER message
2562 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2563 * is passed to if_simloop().
2566 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2568 struct ip *ip = mtod(m, struct ip *);
2572 int datalen = ntohs(ip->ip_len) - iphlen;
2575 /* Keep statistics */
2576 PIMSTAT_INC(pims_rcv_total_msgs);
2577 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2582 if (datalen < PIM_MINLEN) {
2583 PIMSTAT_INC(pims_rcv_tooshort);
2584 CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2585 __func__, datalen, ntohl(ip->ip_src.s_addr));
2587 return (IPPROTO_DONE);
2591 * If the packet is at least as big as a REGISTER, go agead
2592 * and grab the PIM REGISTER header size, to avoid another
2593 * possible m_pullup() later.
2595 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2596 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2598 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2600 * Get the IP and PIM headers in contiguous memory, and
2601 * possibly the PIM REGISTER header.
2603 if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2604 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2605 return (IPPROTO_DONE);
2608 /* m_pullup() may have given us a new mbuf so reset ip. */
2609 ip = mtod(m, struct ip *);
2610 ip_tos = ip->ip_tos;
2612 /* adjust mbuf to point to the PIM header */
2613 m->m_data += iphlen;
2615 pim = mtod(m, struct pim *);
2618 * Validate checksum. If PIM REGISTER, exclude the data packet.
2620 * XXX: some older PIMv2 implementations don't make this distinction,
2621 * so for compatibility reason perform the checksum over part of the
2622 * message, and if error, then over the whole message.
2624 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2625 /* do nothing, checksum okay */
2626 } else if (in_cksum(m, datalen)) {
2627 PIMSTAT_INC(pims_rcv_badsum);
2628 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2630 return (IPPROTO_DONE);
2633 /* PIM version check */
2634 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2635 PIMSTAT_INC(pims_rcv_badversion);
2636 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2637 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2639 return (IPPROTO_DONE);
2642 /* restore mbuf back to the outer IP */
2643 m->m_data -= iphlen;
2646 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2648 * Since this is a REGISTER, we'll make a copy of the register
2649 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2652 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2654 struct ip *encap_ip;
2659 if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2661 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2662 (int)V_reg_vif_num);
2664 return (IPPROTO_DONE);
2666 /* XXX need refcnt? */
2667 vifp = V_viftable[V_reg_vif_num].v_ifp;
2673 if (datalen < PIM_REG_MINLEN) {
2674 PIMSTAT_INC(pims_rcv_tooshort);
2675 PIMSTAT_INC(pims_rcv_badregisters);
2676 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2678 return (IPPROTO_DONE);
2681 reghdr = (u_int32_t *)(pim + 1);
2682 encap_ip = (struct ip *)(reghdr + 1);
2684 CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2685 __func__, ntohl(encap_ip->ip_src.s_addr),
2686 ntohs(encap_ip->ip_len));
2688 /* verify the version number of the inner packet */
2689 if (encap_ip->ip_v != IPVERSION) {
2690 PIMSTAT_INC(pims_rcv_badregisters);
2691 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2693 return (IPPROTO_DONE);
2696 /* verify the inner packet is destined to a mcast group */
2697 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2698 PIMSTAT_INC(pims_rcv_badregisters);
2699 CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2700 ntohl(encap_ip->ip_dst.s_addr));
2702 return (IPPROTO_DONE);
2705 /* If a NULL_REGISTER, pass it to the daemon */
2706 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2707 goto pim_input_to_daemon;
2710 * Copy the TOS from the outer IP header to the inner IP header.
2712 if (encap_ip->ip_tos != ip_tos) {
2713 /* Outer TOS -> inner TOS */
2714 encap_ip->ip_tos = ip_tos;
2715 /* Recompute the inner header checksum. Sigh... */
2717 /* adjust mbuf to point to the inner IP header */
2718 m->m_data += (iphlen + PIM_MINLEN);
2719 m->m_len -= (iphlen + PIM_MINLEN);
2721 encap_ip->ip_sum = 0;
2722 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2724 /* restore mbuf to point back to the outer IP header */
2725 m->m_data -= (iphlen + PIM_MINLEN);
2726 m->m_len += (iphlen + PIM_MINLEN);
2730 * Decapsulate the inner IP packet and loopback to forward it
2731 * as a normal multicast packet. Also, make a copy of the
2732 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2733 * to pass to the daemon later, so it can take the appropriate
2734 * actions (e.g., send back PIM_REGISTER_STOP).
2735 * XXX: here m->m_data points to the outer IP header.
2737 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2739 CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2741 return (IPPROTO_DONE);
2744 /* Keep statistics */
2745 /* XXX: registers_bytes include only the encap. mcast pkt */
2746 PIMSTAT_INC(pims_rcv_registers_msgs);
2747 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2750 * forward the inner ip packet; point m_data at the inner ip.
2752 m_adj(m, iphlen + PIM_MINLEN);
2755 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2757 (u_long)ntohl(encap_ip->ip_src.s_addr),
2758 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2759 (int)V_reg_vif_num);
2761 /* NB: vifp was collected above; can it change on us? */
2762 if_simloop(vifp, m, dst.sin_family, 0);
2764 /* prepare the register head to send to the mrouting daemon */
2768 pim_input_to_daemon:
2770 * Pass the PIM message up to the daemon; if it is a Register message,
2771 * pass the 'head' only up to the daemon. This includes the
2772 * outer IP header, PIM header, PIM-Register header and the
2774 * XXX: the outer IP header pkt size of a Register is not adjust to
2775 * reflect the fact that the inner multicast data is truncated.
2777 return (rip_input(&m, &off, proto));
2781 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2788 if (V_mfchashtbl == NULL) /* XXX unlocked */
2790 error = sysctl_wire_old_buffer(req, 0);
2795 for (i = 0; i < mfchashsize; i++) {
2796 LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2797 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2807 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2808 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2809 "IPv4 Multicast Forwarding Table "
2810 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2813 vnet_mroute_init(const void *unused __unused)
2816 V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2818 V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2819 M_MRTABLE, M_WAITOK|M_ZERO);
2820 V_bw_meter_timers = mallocarray(BW_METER_BUCKETS,
2821 sizeof(*V_bw_meter_timers), M_MRTABLE, M_WAITOK|M_ZERO);
2822 V_bw_upcalls = mallocarray(BW_UPCALLS_MAX, sizeof(*V_bw_upcalls),
2823 M_MRTABLE, M_WAITOK|M_ZERO);
2825 callout_init(&V_expire_upcalls_ch, 1);
2826 callout_init(&V_bw_upcalls_ch, 1);
2827 callout_init(&V_bw_meter_ch, 1);
2830 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2834 vnet_mroute_uninit(const void *unused __unused)
2837 free(V_bw_upcalls, M_MRTABLE);
2838 free(V_bw_meter_timers, M_MRTABLE);
2839 free(V_viftable, M_MRTABLE);
2840 free(V_nexpire, M_MRTABLE);
2844 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2845 vnet_mroute_uninit, NULL);
2848 ip_mroute_modevent(module_t mod, int type, void *unused)
2853 MROUTER_LOCK_INIT();
2855 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2856 if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2857 if (if_detach_event_tag == NULL) {
2858 printf("ip_mroute: unable to register "
2859 "ifnet_departure_event handler\n");
2860 MROUTER_LOCK_DESTROY();
2867 mfchashsize = MFCHASHSIZE;
2868 if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2869 !powerof2(mfchashsize)) {
2870 printf("WARNING: %s not a power of 2; using default\n",
2871 "net.inet.ip.mfchashsize");
2872 mfchashsize = MFCHASHSIZE;
2875 pim_squelch_wholepkt = 0;
2876 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2877 &pim_squelch_wholepkt);
2879 pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2880 if (pim_encap_cookie == NULL) {
2881 printf("ip_mroute: unable to attach pim encap\n");
2884 MROUTER_LOCK_DESTROY();
2888 ip_mcast_src = X_ip_mcast_src;
2889 ip_mforward = X_ip_mforward;
2890 ip_mrouter_done = X_ip_mrouter_done;
2891 ip_mrouter_get = X_ip_mrouter_get;
2892 ip_mrouter_set = X_ip_mrouter_set;
2894 ip_rsvp_force_done = X_ip_rsvp_force_done;
2895 ip_rsvp_vif = X_ip_rsvp_vif;
2897 legal_vif_num = X_legal_vif_num;
2898 mrt_ioctl = X_mrt_ioctl;
2899 rsvp_input_p = X_rsvp_input;
2904 * Typically module unload happens after the user-level
2905 * process has shutdown the kernel services (the check
2906 * below insures someone can't just yank the module out
2907 * from under a running process). But if the module is
2908 * just loaded and then unloaded w/o starting up a user
2909 * process we still need to cleanup.
2912 if (ip_mrouter_cnt != 0) {
2916 ip_mrouter_unloading = 1;
2919 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2921 if (pim_encap_cookie) {
2922 ip_encap_detach(pim_encap_cookie);
2923 pim_encap_cookie = NULL;
2926 ip_mcast_src = NULL;
2928 ip_mrouter_done = NULL;
2929 ip_mrouter_get = NULL;
2930 ip_mrouter_set = NULL;
2932 ip_rsvp_force_done = NULL;
2935 legal_vif_num = NULL;
2937 rsvp_input_p = NULL;
2941 MROUTER_LOCK_DESTROY();
2950 static moduledata_t ip_mroutemod = {
2956 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);